perplexity_skill

Overview

This skill integrates Perplexity AI for web search and conversational research to quickly find up-to-date resources and synthesize web information. It routes queries to the appropriate Perplexity tool (Search or Ask) and enforces safe defaults to keep results focused and concise. Use it only for generic web queries and recent information, not for library docs, workspace internals, or CLI specifics.

How this skill works

The skill inspects the user's phrasing for search triggers like "search", "find", "look up", "ask", "research", or "what's the latest" and chooses the right Perplexity endpoint. For result lists and URLs it runs Perplexity Search with conservative limits (few results, truncated per-result content). For conversational explanations it uses Perplexity Ask to synthesize and explain findings. A prioritized tool chain ensures other specialized MCPs are selected when relevant (library docs, CLI, workspace).

When to use it

• When you want recent articles, tutorials, blog posts, or best-practice guides
• When you ask generic web queries like "search for X" or "find tutorials on Y"
• When you need a concise set of source URLs to explore further
• When you want a conversational synthesis of web information (use Ask)
• When the question is not about in-repo code, framework docs, or specific CLI workflows

Best practices

• Prefer short, focused queries; the skill defaults to limited results to avoid information overload
• If you need comprehensive coverage, explicitly ask to increase result or token limits
• Never use this skill for library/framework documentation — use the dedicated docs MCP instead
• Avoid workspace-specific or CLI-specific queries; route those to workspace/Graphite MCPs
• When deep multi-source research is required, request the Researcher agent instead of Perplexity Research

Example use cases

• Search for the latest React performance optimization articles
• Find tutorials and code samples for setting up PostgreSQL streaming replication
• Look up current best practices for microservices observability
• Ask for a plain-language explanation of advisory locks with recent references
• Find blog posts comparing two competing libraries or services